Inverter DC copper demonstration
This is an executed WayriCAD Quick PI calculation on the exact saved public inverter PCB, not a generated gradient or a measured result. The static image renders values from the solved triangular finite-element mesh. The browser overlay imports the computed data and does not run a solver in the visitor’s browser.
Source and operating conditions
- PCB:
3PD+IB.kicad_pcb, public source revision283f41b65c193abcc7e63c8c919dacdbccc89829. - Exact SHA-256:
09bc4def38e6d866ee38c2ba4f87c349945c3514a4f3b3195c14b89e530fa75d. - Selected net: Vdc, source J10.1, load Q4.2.
- A normalized 1 A constant-current load and 1 V source reference are explicitly assigned. The public README documents a 500 W board but provides no complete source/load operating point. These inputs do not represent the inverter’s operating voltage, output current, or 500 W performance.
- Saved F.Cu and B.Cu thickness: 35 µm each, taken from the board stackup. Copper is modeled at 20 °C, with resistivity 1.724 × 10⁻⁸ Ω m. Plating thickness is an explicit 25 µm assumption.
- The saved pad contact polygons form ideal equipotential electrodes. No additional contact radius or contact resistance is invented.
Model and executed verification
The production Quick PI mesher extracts saved copper tracks, pads, filled zones, holes, and plated barrels. The linear 2.5D DC conductivity FEM solves both copper layers together; the displayed field is F.Cu. Package resistance, semiconductor switching, loaded return paths, AC impedance, temperature feedback, and thermal fields are outside this study.
| Mesh edge | Nodes | Triangles | Source-to-load drop | Copper loss |
|---|---|---|---|---|
| 2 mm | 10,256 | 15,202 | 7.861494 mV | 7.861494 mW |
| 1 mm | 21,786 | 38,262 | 7.911853 mV | 7.911853 mW |
| 0.5 mm | 63,369 | 121,428 | 7.936776 mV | 7.936776 mW |
The final two total-drop changes are 0.6365% and 0.3140%. The final
source/load current mismatch is 1.11 × 10⁻¹⁶ A, maximum nodal residual is
9.88 × 10⁻¹¹ A, and relative power-balance error is 2.19 × 10⁻¹⁶.
An independent 10 × 5 × 0.035 mm rectangular conductor test using the same
production solver agrees with R = ρL/(wt) within a 10⁻⁹ relative tolerance.
These checks support numerical execution and total-drop stability for this declared model. They do not establish physical accuracy or board qualification. The final local peak current-density change is 16.29%, so local hotspots are not mesh converged and are not presented as conductor ratings.
Display mapping
inverter-real-dc.png colors the mean solved voltage drop on every F.Cu triangle
of the final mesh, with a fixed 0–8 mV color scale. Source and load markers use
their saved pad centers.
inverter-real-dc.json contains 424 conservative 2 mm sample squares per field.
Squares crossing copper/hole boundaries are omitted; no values are extrapolated
onto empty board area. Voltage uses barycentric interpolation in an actual FEM
triangle; current density uses the containing element’s solved value. The
reported maxima come from the full solution, not the sampled overlay.
Reproduction
Use an existing Python environment containing KiCad 10 native pcbnew, NumPy,
SciPy, VTK, and matplotlib; no package installation is performed by the script.
From the portfolio repository:
python -B tools/compute-inverter-dc.py --quickpi-root /path/to/WayriCAD
The script preserves the PCB, verifies its exact hash, solves 2/1/0.5 mm meshes,
executes the analytical reference and conservation assertions, and regenerates
the dataset, report, and figure. Per-module solver hashes and runtime versions
are recorded in inverter-real-dc-report.json. The run creates less than 1 MiB
of published assets and requires no build or new environment.